Abstract
The problem of radar-based counting of multiple individuals moving as a single group is addressed using an mm-wave multiple-input-multiple-output (MIMO) frequency-modulated continuous wave (FMCW) radar. This problem is challenging because the different individuals are closer to each other than the range/azimuth resolution, and their bulk Doppler signatures are difficult to distinguish, as they tend to move together. A processing pipeline is proposed, based on the combination of a multiple target tracking algorithm with a classifier to track each group and count the number of people within. Specific salient features are defined for the classifier and extracted from range-azimuth maps and cadence velocity diagrams (CVDs). The proposed pipeline has been experimentally validated in several outdoor scenarios with grouped people. The results show that the combination of tracking algorithm and classifier in the proposed pipeline outperforms alternative methods from the literature as well as a commercial toolbox for people counting.
| Original language | English |
|---|---|
| Pages (from-to) | 20107 - 20119 |
| Number of pages | 13 |
| Journal | IEEE Internet of Things Journal |
| Volume | 10 |
| Issue number | 22 |
| DOIs | |
| Publication status | Published - 2023 |
Bibliographical note
Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-careOtherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.
Keywords
- Feature extraction
- Internet of Things
- Legged locomotion
- mm-wave radar
- People Counting
- Pipelines
- Radar
- radar signal processing
- Radar tracking
- Spectrogram
- tracking and classification